投稿ログイン

MCP Deep Web Research Server (v0.3.0)

@qpd-v

リポジトリ

MCP Deep Web Research Server (v0.3.0) について

Enhanced MCP server for deep web research

基本情報

カテゴリ

データと分析

ライセンス

MIT

ランタイム

node

トランスポート

stdio

公開者

qpd-v

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mcp-DEEPwebresearch": {
      "command": "npx",
      "args": [
        "playwright",
        "install",
        "chromium"
      ]
    }
  }
}

ツール

3

`{ queries: string[], maxParallel?: number }`

maxParallel is limited to 5 to ensure reliable performance

`{ url: string }`

概要

What is MCP Deep Web Research Server?

It is an MCP server that brings real‑time web research into Claude Desktop. It provides intelligent search queuing, enhanced content extraction, and deep research capabilities for users who need structured, sourced information.

How to use MCP Deep Web Research Server?

Install the package globally (npm install -g mcp-deepwebresearch), then install browser dependencies with npx playwright install chromium. Add a deepwebresearch entry to your Claude Desktop config with the command mcp-deepwebresearch. After setup, ask Claude a research question or use the built‑in agentic-research prompt for guided, iterative searching.

Key features of MCP Deep Web Research Server

Intelligent search queue with rate limiting and automatic retries
TF‑IDF relevance scoring and keyword proximity analysis
Google search integration and structured content extraction
Deep research tool with configurable depth and branching
Parallel search of up to five queries simultaneously
Research session tracking with markdown output

Use cases of MCP Deep Web Research Server

Conduct thorough, multi‑source research on a complex topic
Extract and clean content from specific webpages for analysis
Perform parallel searches to quickly gather diverse perspectives
Iteratively refine research direction based on initial findings
Get real‑time, cited information inside Claude Desktop

FAQ from MCP Deep Web Research Server

What runtime does it require?

Node.js >= 18 and Playwright (installed automatically as a dependency).

How can I fix rate‑limiting errors?

Increase the SEARCH_DELAY_MS environment variable or lower MAX_PARALLEL_SEARCHES.

How do I debug issues?

Check Claude Desktop’s MCP logs (e.g., tail -f ~/Library/Logs/Claude/mcp*.log on macOS) or enable debug logging by setting LOG_LEVEL=debug.

What is the default timeout for searches?

The deep_research tool defaults to 55 seconds, and all requests must complete within the 60‑second MCP timeout.

Are there any known limitations?

Parallel searches are capped at 5 concurrent queries, and the server is beta software – some issues may still appear.

コメント

「データと分析」の他のコンテンツ

MCP Simple PubMed

MCP server for searching and querying PubMed medical papers/research database

Salesforce MCP Server

Salesforce MCP Server

mcp-simple-arxiv

Tool to work with arXiv, provide LLM with ability to search and read papers from there

Data Visualization MCP Server

PubMed Analysis MCP Server

A PubMed MCP server.